AB PLC5 PID questions

[Ken Moore]

They are using PD files, so should be somewhat easier than the old non-enhanced processors where everything was integers.
You need to change the engineering units to the scaled span of the temperature inputs.
I see setpoints of 511 or so, but the range is 0-4095 EU, not going to work well.
Also for temp loops, the tuning looks too aggressive, too much gain, and too fast on the integral.

Edit:
The PID is setup for raw counts of 0-4095, make sure the analog card setup matches.

[drbitboy]

Originally Posted by Ken Moore

Also for temp loops, the tuning looks too aggressive, too much gain, and too fast on the integral.

is the Kp (or Kc) divided by 100, so the actual gain is 0.5 and not 50?


Or does the divide-by-100 trick only apply to non-enhanced (that manual is a interesting read, more smoke than light)?

[Peter Nachtwey]

Originally Posted by NetNathan

Yes... PID is so cryptic in plcs that even the manufacture cannot explain well.

The PLC-5 PID design is poor.
The same questions come up over and over again.
Ron Beaufort use to answer all these types of questions.

I despise all the settings required to make a PID work in most plcs, when most temp controller instruments seem to do fine with only setting Proportional band, Derivative (Rate), Integral (Reset).

And some have auto tuning.

[Ken Moore]

The divide by 100 only applies when you are using "N:" type files. With the PD file type the gain value is a float, so you can have a value of 3.4 or 4.25

Once you configure one or two of these, ( I did a bunch, back in the day), then the manual is easier to understand.

I migrated from TI-565's to PLC-5's and was shocked at how crude and clunky PID's and analog handling was in the 5's. The TI platform was far more advanced and easier.

[NetNathan]

Originally Posted by Peter Nachtwey

The PLC-5 PID design is poor.
The same questions come up over and over again.
Ron Beaufort use to answer all these types of questions.




And some have auto tuning.

And very good Auto Tuning at that. The Azbil SDC36 i normally use has a very good and fast auto tune. Also up to 8 PIDS that can assigned in ranges.

[NetNathan]

Originally Posted by drbitboy

is the Kp (or Kc) divided by 100, so the actual gain is 0.5 and not 50?


Or does the divide-by-100 trick only apply to non-enhanced (that manual is a interesting read, more smoke than light)?

I will have to check that. Is it Proportional Band or Gain?
I believe they are inverse of each other...ie..25 PB is 75 gain?

[daba]

Does anyone remember the good old days, when Eurotherm PID controllers were "plug-in" modules that had a 5 or 6 inch drum with a slidewire on it as the SP, and a moving coil galvanometer as the PV display.

I used to service them, long ago, and we always unplugged the I and D sub-boards to disable them, taking care not to disturb the potentiometer settings ! The slidewire got cleaned with a cotton swab and a decent cleaning fluid, and we tested the unit with the I and D modules out, before return to the client with them re-inserted, untouched !

The only reason we got them in for "servicing" was that our client had a policy that if a controller was removed from a machine, for whatever reason, it had to be "serviced" before being put back into use.

For the crudeness of their design, and bear in mind we are going back circa 40 years, they were incredibly good instruments/controllers. I cannot recall any that didn't go back to the client after their "service".

[Ken Moore]

It is GAIN, the larger the number the greater the response to an error.

[NetNathan]

Originally Posted by daba

Does anyone remember the good old days, when Eurotherm PID controllers were "plug-in" modules that had a 5 or 6 inch drum with a slidewire on it as the SP, and a moving coil galvanometer as the PV display.

I used to service them, long ago, and we always unplugged the I and D sub-boards to disable them, taking care not to disturb the potentiometer settings ! The slidewire got cleaned with a cotton swab and a decent cleaning fluid, and we tested the unit with the I and D modules out, before return to the client with them re-inserted, untouched !

The only reason we got them in for "servicing" was that our client had a policy that if a controller was removed from a machine, for whatever reason, it had to be "serviced" before being put back into use.

For the crudeness of their design, and bear in mind we are going back circa 40 years, they were incredibly good instruments/controllers. I cannot recall any that didn't go back to the client after their "service".

Honeywell made a few slide wire controller/recorders also...

[daba]

Originally Posted by NetNathan

Honeywell made a few slide wire controller/recorders also...

It was happy days for us, we got paid for cleaning a slide-wire

[drbitboy]

Originally Posted by NetNathan

I will have to check that. Is it Proportional Band or Gain?
I believe they are inverse of each other...ie..25 PB is 75 gain?

A gain of 75 is equivlaent to a proportional band of 1.33% (=100/75), IIRC. That might not be right, but certainly (gain * proportional band) = some constant, and that constant is probably either 1 or 100%, and the units of that constant are the units of proportional band.

[NetNathan]

Originally Posted by drbitboy

A gain of 75 is equivlaent to a proportional band of 1.33% (=100/75), IIRC. That might not be right, but certainly (gain * proportional band) = some constant, and that constant is probably either 1 or 100%, and the units of that constant are the units of proportional band.

Thanks..

[Peter Nachtwey]

The proportional gain should be 1.33% control output per degree.
The open loop gain ( plant gain ) should have units of degree/ %control output.
However, the Rockwell PIDs seem to do everything in counts.
This is what makes Rockwell PIDs more difficult than necessary.


If I were in charge of Rockwell software I would use the SCP block to convert counts to degrees in floating point format. The output would also use a SCP block to convert the control out% as a real do counts. Now the controller gain has units of %control output / degree C of error.


This would be so much easier.



If the ambient temperature is 25 degrees C and the open loop gain
is 3 degrees C / % control output then an control output of 25% would increase the temperature to 100 degrees C. If you know the plants open loop gain, one should be able to calculate the control output necessary to achieve any temperature that is possible. I don't see why this is so damn difficult yet the same questions have be asked for over 20 years now.

[NetNathan]

Originally Posted by Peter Nachtwey

The proportional gain should be 1.33% control output per degree.
The open loop gain ( plant gain ) should have units of degree/ %control output.
However, the Rockwell PIDs seem to do everything in counts.
This is what makes Rockwell PIDs more difficult than necessary.


If I were in charge of Rockwell software I would use the SCP block to convert counts to degrees in floating point format. The output would also use a SCP block to convert the control out% as a real do counts. Now the controller gain has units of %control output / degree C of error.


This would be so much easier.



If the ambient temperature is 25 degrees C and the open loop gain
is 3 degrees C / % control output then an control output of 25% would increase the temperature to 100 degrees C. If you know the plants open loop gain, one should be able to calculate the control output necessary to achieve any temperature that is possible. I don't see why this is so damn difficult yet the same questions have be asked for over 20 years now.

Thank you..

[drbitboy]

The bang-bang behavior suggests otherwise, and it depends on how the output is scaled, but with the PV range set to [0-4095] and the temperature in degrees, a 1 degree change is 0.025% of full scale, so 50 is not the most outrageous proportional gain.


Or am I still not grokking the scaling of the PD-based PID?